Back to Start Page

The Brahms Sonatas, Op. 120 for Viola and their Textual Challenges

Few works entered as quickly into the violist's literature or have come to occupy as central a position in that repertory as the Two Sonatas, Op. 120 by Johannes Brahms. These compositions were being performed frequently within a few years of their publication and are among the earliest recordings of viola works. To this day, they remain two of the most popular pieces for this noble instrument.

Perhaps as a consequence of its love for this glorious music, the community of violists and teachers usually reacts with consternation when confronted with the true history of the publication and execution of these sonatas. In fact, with the notable historical exceptions of William Primrose and Bruno Giuranna, it seems that most violists to this day prefer to trust dubious publications rather than confront the facts as we now know them. It is the purpose of this paper to present these facts in a useful fashion for those who are interested. It is my hope that we may, someday in the future, hear more performances of these works based upon the critical analysis of available texts rather than upon habit or blind faith in so-called "Urtext" editions.

Prior to publication

The events leading up to the publication of these sonatas are interesting in and of themselves.

Brahms originally composed the Two Sonatas, Op. 120 for his clarinettist friend Richard Mühlfeld of Meiningen, to whom he had previously dedicated the Clarinet Trio and Clarinet Quintet. After finishing the two sonatas, Brahms was requested by the publisher Simrock to create a version for viola and piano.

On October 14, 1894 Brahms wrote to violinist Joseph Joachim, asking whether they could meet in Frankfurt in the first half of Winter. Brahms mentioned that he would try to bring Mühlfeld and possibly a viola part for them to perform for Clara Schumann. Yet Brahms wrote again just 3 days later, stating "I hope that Mühlfeld can come - as I fear I find these two works quite awkward and unpleasant as viola sonatas." This was before he had even created a version for viola!

There is evidence that Brahms sent Simrock a viola part in February or March of 1895, although no such manuscript is known to exist today. It appears that Brahms had little time or interest in this transcription. Indeed, as the publication date, June 1895 (the version for violin was published in July 1895), was approaching, he still had not completed the task. Brahms did, nonetheless—and perhaps out of friendship to Joachim—find time to produce versions for violin and piano which featured not only a complete violin part, but also a substantially revised piano part.

At this point a note about Brahms’ techniques of transcription is in order. In the case of the transcriptions for violin of these sonatas, when a change of register (octave) in the violin part was called for, he also adjusted the piano part accordingly, frequently correcting resulting voice-leading problems. This approach can be observed throughout Brahms’ frequent efforts to transcribe his own music. An alternative approach can be observed in his transcription of the clarinet parts of the Trio Op. 114 and Quintet Op. 115 for Viola. In these cases, he left all the parts as they had been, essentially calling the clarinet part a viola part. He changed no registers and made no other substantial changes. Thus, it appears Brahms’ transcription methods numbered two: adjust all changes in register and correct for voice leading, or adjust nothing at all. And therein lies the first reason to be suspicious of the published version of Op. 120 for viola and piano: the viola part is full of alterations, but no adjustments were made in the piano part.

The conditions leading up to publication of Op. 120 were also unusual. As mentioned, Brahms had not contributed a viola version of the sonatas and the deadline for publication was looming. The publisher, having protested as much as he could, appointed his own staff to produce the transcription. This resulted in the part, evidently prepared by William Kupfer for Simrock. This is the part which can be viewed today in the Brahms Archive in Hamburg, and in which Brahms made some hasty minor corrections before its publication. In further contrast to Brahms’ standard practise, there is no signed viola part from Brahms himself, a fact that should have disturbed publishers and violists alike. I would, however differ with some other opinions on this subject. The copyist’s part (ie. the print-ready part prepared by the publisher) in Hamburg was carefully and conscientiously done. It contains some scrivenor’s mistakes and some mistakes in judgement as well, but is, on the whole, quite clean. Brahms corrections, on the other hand, are amongst the sloppiest I have witnessed. From their appearance, it seems unlikely that he spent more than one hour proofreading the two sonatas. He did correct some errors, but missed others and created additional ones himself. In his manuscripts of the clarinet and violin versions, his work was substantially more careful and exact. Kupfer’s part, hastily corrected by Brahms, remains the principle source for the first and most subsequent editions, including those called “Urtext”.

It is clear at this point that there is no indication that the version for Viola of these sonatas derives from Brahms himself. He had not approved this version for Viola, and the alterations in the subsequent published version for Viola are atypical of Brahms.

There are several aspects related to textual decisions in these sonatas which I will not cover extensively. Matters of taste, as well personal listening and performing habits are best left to the readers to evaluate for themselves. The question of which register sounds best on a specific viola in the hands of a specific performer is also a subjective one best left to others. I have frequently observed experienced performers, when playing alone, to assert that the lower octave sounds better, while reversing their decision, when playing in a large hall with a grand piano.

The question of technical feasibility will also be avoided. While it is clear that Simrock had hoped to sell music to Hausmusiker and hobby violists, and although Brahms was clearly pressured into approving this transcription, there is no way to know why, in each individual case, Brahms accepted particular changes to his musical text. It is clear that an amateur violist in 1895, using a raw gut A-string would have had encountered difficulties unfamiliar to a modern violist with modern technique and strings. One paradoalso remains: why did Brahms not change, in this case, the viola parts in the Trio and Quintet? I leave the reader to resolve this mystery with a few passages for the viola from the Clarinet Trio, approved by Brahms and certainly not less difficult than those in our sonatas.

Example 1

Example 2

I will, therefore, deal with 4 criteria which are more objective, namely:


The viola range surpasses that of the clarinet by two notes (C3 and C3-sharp). There are just three moments where this might have played a role.

  • 120/1, 1st mov´t, m. 66 (C3)
  • 120/1, 4th mov´t, mm. 42-43 (C3)
  • 120/2, 1st mov´t, m. 85 (C#3) (see below, Voice Leading)

Idiomatic writing

There are several examples of idiomatic writing which might speak for one version or another. Double stops and chords certainly fall under this category, being idiomatic for the viola and impossible on the clarinet. These passages, many with chords and double-stops from Brahms himself, are duly noted in AppendiIV.

Not to be excluded are two further deliberations, on which I will elaborate later. I would suggest that any aspiring performer of these sonatas visit the Brahms Archive in Hamburg; examine the clarinet, piano and violin parts, as well as other works of Brahms; and reach his or her own conclusions. Such curiosity and minor investment of time have never yet harmed a performance.

Voice leading

It can be presumed that any scholar familiar with Brahms’ symphonies and chamber music will be familiar with his approach to voice-leading. Therefore, I will point out just a few examples of passages where, in the first and “Urtext“ editions, Brahms original voice-leading has obviously been violated. This presentation is by no means comprehensive, but rather presents a small selection of passages with the original voice-leading indicated. (A comprehensive list of editing issues follows in AppendiIII.)

Op. 120/1, 1., mm. 25 - 30

Example 3

This example, upon examination, demonstrates that the final note of the clarinet arpeggio (mm. 29 & 30) always coincides with the piano’s final quarter note. This is true in mm. 34 and 35. The adjusted viola part misses the mark each time. Why then should the viola not reproduce the original clarinet part?

Op. 120/1, 1st mov’t, mm. 122 – 125

Example 4

Is not the rising line created by these interjections not only more logical but also more exciting in the clarinet version? And doesn’t the imitation in m. 125 then make more sense?

Op. 120/1, 1st mov’t, mm. 139 & 141
As pointed out elsewhere, the A5-flat in m. 139 and the D6-flat in m. 141 are thematic.

Op. 120/1, 1st mov’t, mm.184-188
This is quite clearly a canon at the unison. A canon at the lower octave is obviously not the same thing.

Op. 120/1, 1st mov´t, mm. 187-201

Example 5

As indicated in the original, the clarinet doubles and connects the piano line in these bars. The later version for the viola does not fulfil this function. If the viola is an octave lower, a thematic element is lost (compare G5 – C6 – B-flat5 – A-flat5 with the very first bar of this movement).

For the nonbelievers, those who have not discovered the motivic integrity and remarkable compactness of the thematic material in these sonatas, I will illuminate one of the many wonders of these compositions. As you can see, the first theme of the f-minor sonata is practically identical with the theme of the last movement of the E-flat sonata.

Example 6 (Thematic Comparison)

Op. 120/1, 1st mov’t, mm. 213-220
It is clear in the original that the melodic movement revolves around C5. Whether this is as clear in the lower viola octave (C4) remains open for discussion.

Op. 120/2, 1st mov’t, mm.18-21
The piano’s A5 in m. 17 resolves to B5-flat rather than to B4-flat. Equally obvious is the logic of the B3-flat in m. 20 following the D4-flat rather than a D3-flat.

Op. 120/2, 1st mov’t, mm.38-40

Example 7

Is the voice-leading of this passage not clear, when presented in this fashion?

Op. 120/2, 1st mov’t, m. 85.
This passage is perhaps the most exciting of all, as the viola can render Brahms’ idea better than the clarinet. The descending triplets in m. 82 start with B4, doubling the lower third of the piano. In m. 85, however, they begin with C#5, doubling the piano’s upper line. If it had begun with A#4, as in the viola version, it could not have reached the low C#3 7 notes later. Is not the viola, due to its larger range, an improvement in this case?

Op. 120/2, 1st mov’t, m. 79 – 86

I believe, without wishing to press the point, that the following examples are clear without further explanation.

Melodic shape

Having cast doubt upon the accuracy of the published versions of these sonatas, it is valuable to examine several examples and address the issue of melodic shape. No Brahms scholar needs to be reminded of the importance of intervals in Brahms’ melodic writing. Falling thirds are not interchangeable with rising sixths (Piano Quartet, Op. 60 in a minor, 2nd movement). Nor are fourths and fifths interchangeable in their inversion (Double Concerto, second movement). Few composers dedicated such attention to these aspects of intervallic integrity and melodic shape as Brahms.

It is illuminating to examine several passages of the viola sonatas in this light.

Certainly, when considering melodic shape, any comment about the presumed equivalence of these examples is quickly rendered superfluous.


Already some 70 years ago William Primrose had begun to question the accuracy of the musical texts in question. Bruno Giuranna continued this work and several others have continued to do so. At this time, a major publisher of a so-called Urtext-Edition has even expressed interest in publishing a viola part, based on the original clarinet part, in order to correct previous misunderstandings. This progress would certainly be most welcome. The responsibility lies, however at this moment, with the performers and teachers, who now finally have a chance to evaluate the facts and can, each in their own way, arrive at an edition of these works which respects the parameters of the music. In AppendiIII, each and every variation from the original clarinet part, be it the publisher’s version, Brahms’ corrections, be it the first edition and the “Urtext” are carefully documented. Perhaps with this work, the public can, at some point in the future, have some assurance of hearing these works as Brahms might have intended them to be performed.

Appendix I. Timeline

  • 1885 First performance, Brahms’ 4th Symphony in Meiningen (Bülow)
  • 1885 Friendship with Mühlfeld
  • 1891 Trio and Quintet with Clarinet published (with alternative Viola part)
  • 1894 Sonatas Op. 120 composed and performed with Clarinet (Mühlfeld)
  • 1895:
    • Feb.-March: Brahms sends a manuscript of the viola part to Simrock (lost)
    • March: Brahms asks Simrock for a clean viola part in order to correct it
    • June: Op. 120 published for Clarinet or Viola and Piano. Brahms corrects the Kupfer Viola parts in the last moment.
    • July: Version of Op. 120 for Piano and Violin published

Appendix II: Sources

  • Manuscript for Clarinet and Piano (Brahms-Archiv der Hamburger Staats and Universitätsbibliothek)
  • Clear Copy for Viola and Piano by William Kupfer (corrected by Brahms, Brahms-Archiv)
  • First Edition (Simrock)
  • Urtext Editions
  • Other Editions

Appendix III: Catalogue

A catalogue of the textual alterations regarding the viola part
(VS = viola part, ie. Kupfer/Urtext, KlS = Clarinet part)
(Passages which were lowered an octave an octave and accepted by Brahms are marked with “OCT”)

Sonata Op. 120/1 in f-minor

Allegro Appassionato


mm. 5-12. OCT


mm. 25-35. OCT. Brahms left this passage in the lower octave, creating several voice-leading inconsistencies and obscuring the imitation.


mm. 60-67. Brahms left this rhythmically simplified version in the lower octave, adding only the octave upbeat. The rhythm corresponds to the recapitulation and, of course, the clarinet could not play the low c. Perhaps the viola version is an improvement over that for clarinet?


mm. 79-87. Brahms added the chords in mm. 79 and 86, leaving this in the lower octave. The chords are certainly idiomatic for the viola. The clarinet, also here, could not have played the low c. The voice-leading of this lower version (m. 79) correspond to that in mm. 194 (and consequently also mm. 82, 83, and 86).


mm. 87-89. From the second eighth-note in m. 87, Brahms recorrected these measures into the higher octave.


mm. 92-95. OCT. Brahms left this in the lower octave, altering the voice-leading compared to m. 40.


mm. 123-130. Brahms left m. 125 in the lower octave, adding the chord. The rising line of the canon (d#-g-c#) indicates the sequential resolution in the piano part. The higher clarinet part in m. 129 was the upper voice. Hence, the lower version is a compromise in terms of voice-leading. Brahms did cross out the alto clef in m. 129.


mm. 138-142. Brahms scribbled a new version at the bottom of the page, correcting the copyist’s horrors, but evidently in his hurry he forgot that the a-flat at the end of bar 139 and the d-flat at the end of bar 141 are thematic!


mm. 147-150. Brahms readily accepted the idiomatic double stops added by the copyist.


m. 155. Brahms re-corrected the copyist’s viola part back up an octave to the original register.


Brahms added the octave upbeat to the copyist’s version.


mm. 183-188. OCT. As in mm. 68-73, this is a canon rendered practically incomprehensible by the lower octave in the viola part. Brahms, however, left this as in the copyist’s part, correcting the last note to F4.


mm. 199-200. OCT. Brahms left the lower octave, adding the chordal upbeats in mm. 199 and 201, but also obscuring the voice-leading (compare with mm. 82-83).


mm. 213-220. Brahms here seemingly corrected this passage to the original upper octave and then recorrected it downwards, altering the connecting line in m. 213 and writing “8va basso”. The logic of the upper octave with its circling around the note C, is irrefutable, but any final choice will by necessity have to take sound quality and color into consideration.

Andante un poco Adagio

Neither Brahms nor the copyist made any alterations in this movement. Brahms did, however eliminate the mistaken slurs in mm. 65 and 66.

Allegretto grazioso


mm. 29 and 119. Brahms changed the shape of the figure in these bars. The motivation for this change remains, to my knowledge, incomprehensible.


mm. 35-36 and 124-125. Brahms rewrote these 7 notes in the lower register.



mm. 4-8 and 72-76. Brahms himself stipulated the lower octave in these bars, rendering the voice-leading incomprehensible and the sound less clear.


mm. 42-43. OCT. Brahms left the lower octave in these bars, which does not correspond to the voice leading in mm. 142-145. We shouldn’t forget, however, that the clarinet could not have played the low C of the viola version.


mm. 46-53. Brahms emphatically recorrected the viola part back into the original register


m. 219. The chords, present in some editions, are not original.

Sonata, Op. 120/2

Allegro amabile


mm. 18-21. After the copyist transcribed these bars into the lower octave, Brahms rewrote them in the upper octave on the bottom of the page. Then, indecisively, he scratched these out, adding the chord in m. 18. The lower version creates obvious voice-leading problems (the connection of the piano’s A in m. bar 17 to the B-flat in m.18) as well as inverting the falling third (d-flat – b-flat) in m. 20, one which is of motivic importance.


m. 38. OCT. Brahms left the copyist’s version in the lower octave, destroying the connection between C-A (Clar., m. 38) and D-flat – B-flat (Pno. M. 39).


mm. 40-43. OCT. Brahms left the copyist’s version in the lower octave, interrupting the connection from the piano part and placing the main voice below the accompaniment. He proceeded to write this line an octave higher on a separate page, which he subsequently cancelled.


m. 58. OCT.


 m. 63. Brahms added the chord here.


m. 85. Brahms left the copyist’s suggestion here, with the viola beginning on the lower third of the piano (a-sharp) and using the low c-sharp, unplayable on the clarinet.


mm. 87-92. OCT. Brahms seems to have left the copyist’s lower octave in the viola part here. This creates insurmountable voice-leading difficulties.


m. 92. An “8” is written above the barline.


mm. 113-119. OCT. Brahms, also here, left the copyist’s lower octave, writing a treble clef (m. 117), also “8va” and then cancelling both with “8vo basso”. There can be no question that the lower version creates voice-leading difficulties, an embarrassing inversion in m. 119, and alters the melodic line.


mm. 146-153. OCT. Brahms wrote nothing on this page whatsoever, leaving the copyist to his devices. This of course creates more problems than it resolves, altering the melodic line and obscuring the main voice.

Allegro appassionato


mm. 119-121. It appears that Brahms added or at least corrected the c-sharp – c-double sharp in these bars, as well as the d-sharp in m. 121.


mm. 126-135. The copyist added chords and double stops here, extending the viola part through m. 135, up to the pp in the piano part. He also, unfortunately eliminated the charming and lilting dotted rhythm, the only one of this movement (mm. 126-7), from the clarinet part. Brahms made only minor corrections here.

Andante con moto


mm. 46-47 & 53. OCT. Brahms left the copyist’s lower octaves here, to little effect. In fact, the last three tones of bar 53 are thematic (as in m. 11)


mm. 74-76. Some editions have these bars one octave higher. This is however not original.


mm. 84-85. The octaves in the viola part seem to have been added by Brahms.


mm. 136-38. OCT. Brahms left here the copyist’s lower octave. This is a canon, which is hardly perceptible in the lower octave (see Voice Leading, above). In fact, in the lower register, the viola would sink below the bass in the piano.


mm. 143-147. Brahms seems to have corrected m. 143 downward, eliminating however the subsequent characteristic two-octave jump. The gestual motion of the original seems more suited to this delightful and joking moment, particularly in m. 146.

Appendix IV: Idiomatic alterations


Op. 120/1, 1st mov’t., m. 79


Op. 120/1, 1., m. 86


Op.120/1, 1., m. 125 (Ex. 4)


Op. 120/1, 1., mm.147-150


Op. 120/1, 1., m.175


Op. 120/1, 1., m. 194


Op.120/1, 1., m. 199 (Ex. 5)


Op. 120/1, 1., m. 201 (Ex. 5)


Op. 120/2, 1., m. 18


Op. 120/2, 1., m. 63


Op. 120/2, 2., m. 121


Op. 120/2, 2., mm. 126-135


Op. 120/2, 3., mm. 84-85

Copyright: James Creitz, 2003

Back to Top